The Part That Stays Human

A new study from Anthropic asks a plain question, which jobs can today's robots actually do, and the answer lands close to home for families. Robots can already do most physical work somewhere, the authors find, but mostly in places built for them. Out in the wider world they do "only 2% of physical tasks in unstructured environments." A home is about as unstructured as it gets.

Care work sits at the far end of the scale. Personal care jobs are among the least exposed, because they depend on in-person social interaction and physical contact with humans, "which both LLMs and robots struggle with." Healthcare support scores low too. And even where a robot could do the task, people may not want it to. The authors' example of work held back by trust is a robot asked to "Dress children and change diapers."

Then there is cost. Robots are cheaper than a person for just 0.3 percent of work today. If robot prices keep falling at their past pace, the authors estimate it would take around 40 years for that share to reach 10 percent. New humanoids could change the math, they allow, but nothing in today's numbers says it is close.

For a family, that reads as a timeline. The bath, the transfer out of bed, the pills handed over with a glass of water, the conversation that tells you something is off. That is the work this study says stays human the longest, and it is the work family caregivers do every day, most of it without pay.

So the useful technology for the next decade is not the machine that replaces those hands. It is the tool that makes them less alone, the reminder, the shared list, the alert, the plan. The authors close by wondering whether work becomes "more social and interpersonal as robots and AI advance." For caregivers, it already is.